Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932712Ab0KQNFX (ORCPT ); Wed, 17 Nov 2010 08:05:23 -0500 Received: from mail-bw0-f46.google.com ([209.85.214.46]:43847 "EHLO mail-bw0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754979Ab0KQNFV convert rfc822-to-8bit (ORCPT ); Wed, 17 Nov 2010 08:05:21 -0500 DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=date:from:to:cc:subject:message-id:in-reply-to:references:x-mailer :mime-version:content-type:content-transfer-encoding; b=nmVUZG4jmXd8A4XdMloI7Jqj5DAVsuhdGCHLgjUzLYePUpBNIQCl2QpfgHlDQLfM+Z FDv5uNL+pmGC3boIi13T7lkzvBy/RWKQOKENeM5paCaaG6Al+YC8o3sUsDN/TfVCirEj UXCudUg9MbrnQvaLOIyAwzfEshB9OZmB585Ik= Date: Wed, 17 Nov 2010 15:05:16 +0200 From: =?UTF-8?B?VMO2csO2aw==?= Edwin To: Peter Zijlstra Cc: Thomas Gleixner , LKML , Linus Torvalds , Andrew Morton , Ingo Molnar , Steven Rostedt , Arjan van de Ven Subject: Re: [ANNOUNCE] New utility: 'trace' Message-ID: <20101117150516.18136834@deb0> In-Reply-To: <1289998312.2109.745.camel@laptop> References: <20101117144758.38bf4f05@deb0> <1289998312.2109.745.camel@laptop> X-Mailer: Claws Mail 3.7.6 (GTK+ 2.20.1; x86_64-pc-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1320 Lines: 34 On Wed, 17 Nov 2010 13:51:52 +0100 Peter Zijlstra wrote: > On Wed, 2010-11-17 at 14:47 +0200, Török Edwin wrote: > > I see --call-graph in the trace record -h output, but it doesn't > > seem to work on x86_64 for me. > > If you want to unwind userspace you need to build everything with > framepointers -- sadly some people have been lobbying to remove > framepointers from all distro builds, even though on x86_64 its nearly > free (i386 does have a significant performance benefit, sadly). > > There's some work on-going to simply copy out the top of stack and let > dwarves go wild at it in userspace, but that isn't done yet. Even without frame-pointers I should see at least the first caller in userspace, since that can be obtained by reading the return address of the syscall, right? I don't see any difference in the output of 'trace report' thats why I'm asking it is implemented at all right now, or I need additional flags to see it. I do see some difference in output of 'perf record -g/perf report' though. Best regards, --Edwin -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/